I've been looking into Monte Carlo methods for valuing american options. Now, I found an R code by Stefano M. Iacus that values the option using a tree (based on Broadie and Glassermann) without use of antithetic branching. My goal is to modify the code to include antithetic branching for variance reduction purposes. I've successfully simulated the tree using antithetic branching, but I'm having trouble with the altered lower estimator. See the modified code below, as far as I can see I have coded the low estimator correctly, but when I run the code I get a lower estimator that always equals the upper estimator - I have run the code for many different stocks. My question: Is there even a problem, or have I missed something in the code, if yes, then what? Thank you very much, regards.
(f is a function calculating the immediate payoff at a node) (b is branches)